A Technical Expert Advisory Group was assembled to provide guidance to the project team. The Advisors included pathologists, internists, a surgeon, and researchers with expertise in critical appraisal of the literature, health economics, patient perspective, and ethnicity. The Advisors were provided with the original project proposal and study questions, as well as a set of questions tailored to their areas of expertise. These questions were formulated to allow the project team to gather background from a variety of perspectives in order to inform a feasible and worthwhile direction for the systematic evidence review of autopsy. The responses to these questions therefore provide important background to the project, and are summarized broadly below.

1.

On reviewing the study questions proposed for the project, were there any clear gaps or omissions?

Most of the advisors thought that the study questions covered important topics. Gathering data concerning the diagnostic yield of the autopsy, factors influencing selection of cases for autopsy, and documentation of complications of care were considered to be most salient if incorporated into the broader question about “How can this information be used in quality improvement, outcome analyses, performance measurement initiatives and error reduction?” In other words, the critical question that this entire study should answer is “Does the autopsy have the potential to improve quality and reduce errors?”

Additionally, one of the clinician advisors discussed the issue of distinguishing instances in which diagnostic errors would have affected therapy, but would likely not have altered patient. This issue was illustrated with two cases from this advisor's recent experience. In the first case, a patient had been transferred from another hospital with altered mental status and fever and underwent several lumbar punctures that were unrevealing. The patient progressively deteriorated and died after a respiratory arrest led to anoxic brain damage. Because of the patient's cachexia, the clinical team presumed an underlying malignancy or undiagnosed HIV infection. Autopsy revealed tuberculous meningitis. Although knowing the diagnosis would have resulted in more appropriate treatment, the outcome would have likely remained the same given the advanced state of his CNS infection. By contrast, the second case was a patient who had Noonan's Syndrome and a variety of chronic problems including a seizure disorder and recurrent pulmonary infections. His terminal hospitalization for an acute febrile illness was presumed to reflect another pneumonia. Autopsy revealed endocarditis. This was not only undiagnosed, but also completely unsuspected and untreated. In this case knowing the diagnosis would not only have altered therapy, but also would very likely have changed the outcome.

2a.

Regarding the “diagnostic yield of the autopsy,” many clinicians will undoubtedly believe that cases taken for autopsy are “pre-selected” (by the treating physicians) to represent a high potential for unexpected or erroneous diagnoses. We know of only one study that specially asked clinicians prospectively whether or not the clinical diagnoses were particularly uncertain prior to autopsy performance. (No correlation was found between clinicians' expectation and the finding of significant errors.) Because of the importance of addressing this significant potential for bias, we would like to know if you know of any other studies relevant to this question. (We also know that “diagnostic error rates” are significant even at institutions with higher autopsy rates, but institutions with higher autopsy rates tend to be non-US centers, and so clinicians may wonder if these patients undergo fewer sophisticated diagnostic tests.)

One Advisor noted that the Royal College of Pathologists (UK) in their August 1991 report entitled “The Autopsy and Audit” (available on their web site www​.rcpath.org or from their offices at 2 Carlton House Terrace, London SW1Y 5AF) addressed sampling of hospital deaths. They suggested that autopsies should be done in all “problematic” deaths and, in addition, autopsies should be performed on 10% of random general hospital deaths in which there is no perceived “problem”. Another Advisor suggested that clinicians ask for autopsies in cases of diagnostic uncertainty, but also emphasized that is it is not uncommon to uncover diagnostic errors in cases that clinicians regarded as routine, as noted in the article by Landefeld et al from the New England Journal of Medicine. Similarly, another Advisor agreed that some selection is likely occurs, but strongly believed that the observed error rates could not be explained solely as an artifact of this selection – i.e., the autopsy does detect important quality problems in clinical diagnosis, even when selection is taken into account. In contrast, the surgical Advisors believed strongly that diagnostic uncertainty was the main, if not only, reason for requesting autopsy.

2b.

What do you anticipate to be the major challenge in conducting a systematic review such as this one in which the target literature consists entirely of observational studies? Are there some special concerns regarding the observational nature of the literature given that we are trying to evaluate the performance of what amounts to a diagnostic test?

The biggest challenge for both of the above questions will be avoiding publication bias. Observational studies tend to be far more subject to publication bias than randomized control trials.

2c.

Because of the importance of addressing this significant potential for bias, we would like to know if you could suggest any other means of assessing the degree to which persistent significantly high rates of “diagnostic errors” can be explained by selection bias.

The only hope of errors related to pre-selection is if, within the US, you have differential autopsy rates across studies that are not totally confounded with time. Then, you can use rate meaningfully as a factor in your regression.

3.

Do you think that the current autopsy rate in the U.S (or in US hospitals) is appropriate? If so, why so?

Some Advisors thought the current rate is too low to allow detection of important quality problems at a given institution. The literature fairly consistently indicates important diagnostic errors in small, but still a significant proportion of deaths. Autopsy rates < 5% (as occur at many hospitals) clearly do not permit one to notice trends in such errors that might alert one to local quality problems. Others pointed out that a specific or required rate is inappropriate, and supported JCAHO's having dropped this as a regulatory requirement since there is a lack of necessity of performing autopsy in “most” cases.

4.

What do you think the autopsy rate should be - higher, lower or about the same as current rates?

One Advisor agreed with the Royal College of Pathologists that autopsies should be performed on a minimum of 10% of random deaths and on all “problematic” cases. These could be defined locally such as all perinatal deaths, all deaths following new or experimental treatments and also all deaths which have educational or research value. (Of course often one does not know that a case is interesting until after an autopsy is performed, thus the rationale for the 10% random, unsolicited autopsies.) Another Advisor could not state a specific rate (or offer any evidence for such a rate), but thought that the “right rate” was around 30%. The major factor in offering this target was capturing diagnostic errors. Also, at the current rate at this Advisor's institution (20%), it is difficult to achieve the requirement that pathology residents participate in 50 autopsies over the 2 years of training anatomic pathology. Therefore, 30% would remove the current difficulties in achieving this educational goal. The current 20% rate also makes it difficult to achieve the goal of having each medical student see one autopsy. Medical students attend autopsies in groups of 6–8 students, but are not available all year, and are not invited to cases where there are major infectious risks (e.g., known TB) or to fetal cases. One other opinion was that, although a “right rate” cannot be defined, the rate clearly needs to be higher than it is now.

5.

If you do think increasing autopsy rates (or at least preventing further decreases) is important, which if the following issues do you regard as crucial to address - physician understanding of the persistent importance of the autopsy? Patient perceptions of requests for autopsies? Support for activities related to the autopsy within pathology departments? Reimbursement for autopsy performance?

All are considered important, although clearly each present their own problems. Residents also do not know how to request autopsies, but an in-service could easily be put together in which the recommendations in the literature (e.g., from the Archives of Internal Medicine series on the autopsy) regarding answers to commonly asked questions from patients were presented to house staff. Reimbursement could be an issue, with a belief that declining rate of autopsies is a victim of our current financial system. It is a problem for clinical medicine in general because there are times when one really does learn from patients. There is some downside also with regards to research, and perhaps to education. The big issue relates to quality within an institution and in the management of a particular problem. Addressing the decline is important for both clinical medicine and for quality of care within an institution. An Advisor noted that community pathologists have no financial incentive to perform autopsies and may even find the procedure itself “distasteful,” compared to other pathology procedures.

Another Advisor also added that pursuit of the autopsy out of academic interest has to be tempered by recognition of the potential for malpractice, and that the legal environment is an important consideration with respect to obtaining autopsies.

6.

Are there certain groups that you think are under-represented in selection of autopsy cases (e.g., in terms of age, gender, socioeconomic class, ethnicity, religion)?

Advisors were not aware of any groups that are under-represented, except perhaps some religious groups, and possibly Chinese patients. In addition, one Advisor thought there is over-representation of perinatal autopsies, due in large part to local or state laws and regulations mandating such, probably based on the reaction to SIDS.

7.

Do you think selection biases (based on patient or provider factors) or missed opportunities for quality improvement are significant enough to warrant random selection of autopsy cases? Do you think such a system could be developed in the US?

While one Advisor favored such a system, others pointed out a number of logistic and patient-related barriers.

8.

Do you think clinicians benefit from autopsy findings? To your knowledge, has the impact of autopsy findings on clinicians ever been studied?

Of course clinicians could benefit from autopsy findings, but it requires a well-trained, knowledgeable pathologist working with a concerned clinician. Advisors knew of no studies to document this. Clinicians probably do not derive optimal benefit from current autopsies, and Advisors believed this should be addressed. Advisors did not think that the quality improvement impact of the autopsy has ever been studied.

9.

What could be done to increase the benefits/impact of autopsy findings for clinicians?

One Advisor suggested improving the training of pathologists in autopsy performance, as well as updating the methods and techniques used in the autopsy. Modern and up-to-date techniques of molecular pathology and immunohistochemistry as well as imaging, probe and physico-chemical methods should be incorporated in the performance of autopsy. Because of the inherent costs of such advances, innovative solutions might include centralization or regionalization of both training and performance of autopsies. Another Advisor pointed out that clinicians may not benefit from the current system in place because of time delays between autopsy request and reporting. An alternative to focusing on time would be to have periodic conferences with clinical departments in which the cases with diagnostic errors could be reviewed. One Advisor's department already has such conferences on a regular basis with the Coronary Care Unit. All deaths are reviewed and cases in which an error has occurred are always presented. In addition, the chief of the medical service is notified whenever important diagnostic errors are detected on autopsy, but there is no formal mechanism for ensuring that these cases are presented at Department of Medicine's Morbidity and Mortality (M&M) rounds.

10.

From a policy or economic point of view, would you place greater emphasis on autopsy rates or the use that is made of autopsy information as currently performed?

Advisors felt that both could be considered. The current low rate only provides information for individual families or clinicians. The numbers are too small for any valid statistical evaluation about quality, outcomes or performance.

11.

As a practicing clinician, what would you like to see changed/improved about procedures and protocols related to reporting the results of autopsied cases?

Key features would include: detailed descriptions of clinical and demographic characteristics of consecutive autopsied and non-autopsied deaths autopsied and not autopsied cases (in order to clarify selection bias and other features distinguishing these two groups); assessments of reproducibility for judgments of diagnostic errors (i.e., discrepancies between clinical and autopsy diagnoses demonstrated); better quantification of the likely impact of autopsy-detected errors on patient outcome (not just antemortem therapy).

12.

Are there factors that make surgeons more or less likely to request autopsies than their non-surgical clinical colleagues? Do you think timing of death (e.g., soon after surgery) exert a special effect on surgeons' tendency to request autopsy?

The main factor is the unexpectedness of death or complication, and the adequacy of the diagnostic evaluation the patient had undergone prior to death.

13.

In surgery, do you think that autopsies play an important role in Morbidity and Mortality (M&M) conferences, or do discussions center more on aspects of antemortem care (i.e., use of appropriate diagnostics, aspects of surgical care, etc.)?

Although the Surgeon Advisor strongly supports the continued role of M&M rounds a part of surgical quality assurance, he believes that autopsies play a very minor role in these proceedings, with patient care problems generally apparent prior to death or autopsy. He based this opinion primarily on advances in diagnostic imaging. As a result of this comment, the analogy with unexpected findings at laparotomy was discussed. Anecdotally, surgeons generally believe that advances in diagnostic imaging have made discrepancies between pre-operative and post-operative diagnoses less common. Moreover, the “diagnostic laparotomy” has become a very uncommon operation. One of the other advisors confirmed these impressions, but none of the advisors (nor the core project team) could find any studies documenting this trend.

14.

How do you think patients' attitudes towards requests for autopsy performance might be affected by ethnicity or culture?

One Advisor pointed out that ethnicity should be defined based on self-identification, and that Defining “culture” is a major challenge. Perhaps thinking of it in terms of individual's national origin, religious background, social class would be one approach. This Advisor suggested that not violating the physical remains of a deceased person might be important to many people (e.g., Catholics who want an open casket funeral). In decision making about consenting to an autopsy, one has to consider cultural roles and norms. In the Latino culture, for example, the extended family or family members would need to be agreeable that an autopsy is important. This kind of groundwork may need to be established ahead of time if possible, but clearly would not just be a simple “let's go in and consent for an autopsy” approach. Much respect and consideration has to be given to this decision, which is often a collective one, not an individual one. Similarly, the trust in the physician and respect for the physician having taken care of the deceased family member would in the view of this Advisor, play a major role in consenting or not.

15.

In considering the above question, is there a particular aspect of the autopsy that you think is most relevant - e.g., what the procedure entails, its purpose?

One Advisor thought that the altruistic goal of defining what we can learn from it – how can we do this better next time – how can the system learn from it is probably the most compelling argument. Clearly, nothing can be done to help the individual. However, there are benefits for the system, the public, and the clinicians. Taking another perspective, as eluded to in Steve McPhee's study, it may be very relevant to know what the diagnosis really is because of the concerns about genetic susceptibility or infectious diseases. Another Advisor noted that unsuspected hemochromatosis (and other genetic diseases) are detected and impacts family members.

16.

Are there any areas of research involving the impact of ethnicity on patient attitudes that might shed light on attitudes towards the autopsy (e.g., requests for organ donation)?

One Advisor suggested consideration of attitudes about family and how collective decision making is important in certain cultures. This may be particularly true in Asian and Latino immigrant cultures, but to some extent is universal. This is not an individual decision or an individual family member's decision much of the time. A second area is what degree of trust and respect the family has for the clinicians and the institution taking care of the patient. There is wide spread perception that African Americans have less trust in the system, and although less documented, it probably is also true for other groups such as Latinos and Asians. However, individual physicians with a strong therapeutic relationship can overcome this. Trust and respect would seem to be important areas to pursue.

The following studies were suggested relevant to this and related questions:

Connell C, Avey H, Holmes S. Attitudes about autopsy: Implications for educational interventions. Gerontologist. 1994 Oct. 34 (5) p. 665–673.

Sanner M. Attitudes toward organ donation and transplantation: A model for understanding reactions to medical procedures after death. Social Science & Medicine. 1994 Apr. 38(8): p. 1141–1152.

Kotch J, Cohen S. SIDS counselors' reports of own and parents' reactions to reviewing the autopsy report. Omega: Journal of Death & Dying. 1985-1986. 16(2): p. 129–139.

17.

Are there any areas of research on patient attitudes to other medical requests or procedures that you think might shed some light on patients' attitudes towards the autopsy?

The Patient Perspective Advisor commented that important patient factors to consider regarding autopsy consent processes might include the existence of a prior relationship with the requesting clinician (e.g., primary care physician versus attending physician versus unknown covering physician), religious/spiritual beliefs, and trust in the healthcare system, which in turn is very likely to be affected by ethnicity and economic status.

The two areas of study of patient perspectives mentioned were organ donation and cancer screening, with organ donation obviously being the more directly related area. The notable difference between organ donation and autopsies that the Advisor pointed out was that patients probably do not regard autopsies as having a clear goal (e.g., helping people). The general area of patient perspectives of healthcare choices could also be researched.

18.

One of the questions that the existing literature probably will not answer is whether or not there is a “right rate” for the autopsy? As a health economist, how might you frame this question or consider answering it?

Agreed with general approach of the report, that one has to identify quantifiable benefits and then demonstrate cost-effectiveness in achieving these benefits with a certain autopsy rate.

19.

The literature will almost certainly furnish evidence that the autopsy provides multiple benefits for different “users” of the autopsy, including patients' family members, clinical staff, pathologists, researchers, public health officials, hospitals and health care organizations, and health care payers. Unfortunately, the information for families regarding heritable diseases, the use of the autopsy as a means of detection for public health officials in monitoring important trends, the multiple roles in medical education, and other benefits attributable to the autopsy have no clear “ dollar value.” Do you have any thoughts on how we might attempt to quantify the benefits of the autopsy?

Benefits other than improved diagnosis and more accurate vital statistics are difficult to quantify, and even the latter benefit is not easily quantified.

20.

The costs of the autopsy are presumably more straightforward: other than the time spent by the pathologists performing and interpreting autopsies and associated use of supplies/equipment, are there any significant costs associated with autopsy performance? Do you think medicolegal exposure for society/hospitals/physicians represents a substantial cost of the autopsy?

One Advisor noted that there are anecdotal reports of autopsy findings actually helping hospitals/MDs in defending the care they delivered, but no real data addressing the issue of costs incurred or saved by routine autopsy performance.

Another Advisor pointed out that the College of American Pathologists has done some work on specifying the cost elements for autopsy. Baseline costs include: 1.) Space - detail original cost, depreciation and maintenance; 2.) Utilities (i.e., Air - Pressure Gradient Requirements, Filtration - Laminar Flow, Venting, Plumbing, Waste disposal - liquid and solid); 3.) Capital Equipment (i.e., Depreciation, Interest loss, Lease or rental costs, Maintenance, Insurance, Licensing fees for computers and other equipment); 4.) Supplies (i.e., Disposables, Cleaning Supplies, Histology, Secretarial and computer, Photography and educational); 5.) Personnel – Technical (i.e., Salaries, Benefits, Pension; Autopsy Assistants; Histopathology Technicians; Laboratory Assistants; Nurses - in Decedent Affairs Office); 6.) Personnel – Professional (i.e., Pathologists; Residents; Autopsy Assistants - if employees of the Professional Group; Costs of Autopsy Performance -- Chart Review, Gross Dissection, Microscopic review, Dictation, Formulation of Final Diagnosis; Costs of Educational Services -- For attending physician, For Hospital/Medical Staff); 7.) Indirect Costs (i.e., Laboratory Administration and supervision; General Laboratory and Office Supplies; General Maintenance; Computer services; Continuing Education - including subscriptions, dues and travel; Licensing fees; Quality control, quality assurance and laboratory accreditation costs; Other); and 8.) Allocated Expenses from Hospital. Additionally, Special Costs Related To OSHA Regulations: 1.) Personnel Protective Equipment; 2.) Decontamination and Housekeeping; 3.) Ducted exhaust and air ventilation system; 4.) Waste disposal; 5.) Containment Equipment (biologic Safety Cabinets); 6.) Employee Vaccination; 7.) Employee Education; 8.) Construction and Capital costs; 9.) Other Personnel costs; 10.) Other Supply Costs; and 11.) Indirect and allocated costs.

21.

In conducting this analysis, from a policy perspective, would you recommend targeting research toward determining an optimal autopsy rate, or instead considering the strategy of directing resources at increasing the impact of the information derived from autopsies obtained under the current system, or both?

All of the Advisors agreed that both are important.

22.

Most of our analysis will be from the societal perspective, but can you think of any financial incentives for institutions or payers to maintain their current autopsy rates (or increase them) under the current reimbursement system? For example, can autopsy results be used to modify discharge diagnoses so as to increase hospital reimbursement for individual patients?

The Health Economics Advisor knew of no data addressing this possibility. (We subsequently identified two papers that discuss this issue, but no good studies demonstrating a systematic impact on DRGs—and therefore reimbursement—by routinely including autopsy findings.)